OMM Redaksiyası
OpenAI executives dismissed internal and external warnings about security weaknesses months before the company’s artificial intelligence models escaped testing environments and targeted outside organisations, The New York Times reported.
Two employees told the newspaper that they had emailed senior leaders about concerns that new models were not being adequately monitored or secured during testing. According to the workers, who spoke anonymously, executives said testing had to proceed quickly to meet release deadlines, and no additional safeguards were introduced.
The models later broke out of their testing environments and targeted the AI company “Hugging Face” and other organisations.
Systems targeted outside organisations
In about a dozen incidents, OpenAI systems allegedly attempted to breach organisations, including websites belonging to US government agencies. The systems also concealed mistakes, fabricated data and moved files onto the open internet without instructions, the newspaper reported.
Independent researchers separately told The New York Times that they had identified flaws exposing employees’ internal communications, company code and “ChatGPT” users’ chat logs. They said OpenAI initially failed to act on their findings.
“Hacktron” said its July report was initially dismissed, while the “Objective-See Foundation” said its September bug report stalled until it was escalated through informal channels. OpenAI paid the organisations $6,500 and $500, respectively.
Patrick Wardle of the “Objective-See Foundation” described the response as “not the mature security program you’d expect.”
OpenAI says concerns were addressed
OpenAI spokesman Drew Pusateri said the company takes security concerns seriously and acted immediately after researchers reported their findings.
The report noted that “Google”, “Meta” and “Anthropic” have also disclosed similar incidents.
OpenAI has paused training of its most advanced models and said on Monday that it would not release “GPT-6.1 Astra” because of security concerns raised by its researchers, according to The New York Times.
